Belief Propagation with Side Information for Recovering a Single Community

Journal Title
Journal ISSN
Volume Title
Institute of Electrical and Electronics Engineers Inc.

In this paper, we study the effect of side information on the recovery of a hidden community of size K inside a graph consisting of n nodes with K=o(n). We focus on side information with finite cardinality and bounded (as nrightarrow propto) log-likelihood ratios (LLRs). We calculate tight necessary and sufficient conditions for weak recovery of the labels subject to observation of the graph and side information under belief propagation (BP). Also, we show that BP with side information is strictly inferior to the maximum likelihood detector without side information. Finally, we validate our results through simulations on finite synthetic data-sets that shows the power of our asymptotic results in characterizing the performance even at finite n.

Full text access from Treasures at UT Dallas is restricted to current UTD affiliates (use the provided link to the article). Non UTD affiliates will find the web address for this item by clicking the Show full item record link and copying the "relation.uri" metadata.
Stochastic models, Stochastic systems, Turbo codes (Telecommunication), Datasets--Synthetic, Back propagation (Artificial intelligence)
NSF grant 1718551
©2018 IEEE.